@grin
ppl often ask me how to make a bot like @neynar. my answer is usually some combination of high-level philosophy and very basic template:
philosophy -- its kinda like raising kids
you ship the bot and then you spend every day lovingly paying attention to it and caring for it little by little. there's no shortcut. the latest "best practices" change all the time and its more vibes than science. that said, it helps if you really care about what you're making and you have a clear purpose in mind. also helps to get feedback from other ppl who interact with your creation.
template -- i would start super simple:
1. neynar webhook that hits your api anytime the bot is mentioned
2. api call to llm provider with your prompt + the contents of the cast. i use Vercel's AI SDK for this step, but you could do @elizaos or another framework. if this is a long-term project, pick something simple that handles the basics and is not too opinionated.
3. publish the response via neynar api or directly via snapchain
then once you've started and you have specific challenges to overcome, i'll have much more tactical advice for you